lichess.org
Donate

elo and dwz

I have taken the tables of the Grenke A, B and C tournaments and calculated the linear dependency between ELO and DWZ (Deutsche Wertungszahl, german rating number). For the A and B categories, the correlation is quite high. Not so for the C category.

Here are the results

                      Formula  =>   DWZ = a * ELO + b
Class   n     Mean DWZ    Mean Elo    Slope (a)   Intercept (b)   Pearson r   R^2         approx. DWZ Range
A       602   2135.6495   2139.1362   0.932880    146.831335      0.959376    0.920403       2000-2500
B       1010  1660.6950   1762.4703   0.514670    907.759979      0.839404    0.704600       1400-2000
C       106   1166.7830   1513.0849   0.159700    1326.749467     0.446591    0.199443       1000-1600

By the way the title of this post should read ELO and DWZ, but it seems Lichess changes caps to lowercase in titles.

I have taken the tables of the Grenke A, B and C tournaments and calculated the linear dependency between ELO and DWZ (Deutsche Wertungszahl, german rating number). For the A and B categories, the correlation is quite high. Not so for the C category. Here are the results ``` Formula => DWZ = a * ELO + b Class n Mean DWZ Mean Elo Slope (a) Intercept (b) Pearson r R^2 approx. DWZ Range A 602 2135.6495 2139.1362 0.932880 146.831335 0.959376 0.920403 2000-2500 B 1010 1660.6950 1762.4703 0.514670 907.759979 0.839404 0.704600 1400-2000 C 106 1166.7830 1513.0849 0.159700 1326.749467 0.446591 0.199443 1000-1600 ``` By the way the title of this post should read ELO and DWZ, but it seems Lichess changes caps to lowercase in titles.

The discrepancy in the C category is likely due to the FIDE rating adaption last year. Before that DWZ and Elo were quite close, but due to the adaption DWZ 1000 corresponds to roughly FIDE Elo 1400.

The discrepancy in the C category is likely due to the FIDE rating adaption last year. Before that DWZ and Elo were quite close, but due to the adaption DWZ 1000 corresponds to roughly FIDE Elo 1400.

@FirstRaven said ^

The discrepancy in the C category is likely due to the FIDE rating adaption last year. Before that DWZ and Elo were quite close, but due to the adaption DWZ 1000 corresponds to roughly FIDE Elo 1400.

The transformation was linear so that it would only change the resulting coefficients. IMHO the reason rather is that most lower rated players do not have many FIDE rated games so that their FIDE rating is less reliable and more prone to random fluctuations.

@FirstRaven said [^](/forum/redirect/post/lS1Vvqnm) > The discrepancy in the C category is likely due to the FIDE rating adaption last year. Before that DWZ and Elo were quite close, but due to the adaption DWZ 1000 corresponds to roughly FIDE Elo 1400. The transformation was linear so that it would only change the resulting coefficients. IMHO the reason rather is that most lower rated players do not have many FIDE rated games so that their FIDE rating is less reliable and more prone to random fluctuations.

@Alakaluf said ^

By the way the title of this post should read ELO and DWZ, but it seems Lichess changes caps to lowercase in titles.

No, Elo should read Elo (it's a guy's last name).

@Alakaluf said [^](/forum/redirect/post/vsRRJPkZ) > By the way the title of this post should read ELO and DWZ, but it seems Lichess changes caps to lowercase in titles. No, Elo should read Elo (it's a guy's last name).

My first post has got an error, but I cannot edit it anymore. Sorry. The table should read like

                      Formula  =>   ELO = a * DWZ + b
Class   n     Mean DWZ    Mean Elo    Slope (a)   Intercept (b)   Pearson r   R^2       approx. DWZ Range
A       602   2135.6495   2139.1362   0.932880    146.831335      0.959376    0.920403       2000-2500
B       1010  1660.6950   1762.4703   0.514670    907.759979      0.839404    0.704600       1400-2000
C       106   1166.7830   1513.0849   0.159700    1326.749467     0.446591    0.199443       1000-1600

       Inverted =>  DWZ = a' * ELO + b'
Class  Slope(a')   Intercept(b')  
A      1.072146     -157.460         
B      1.943044    -1765.019       
C      6.262484    -8306.006
My first post has got an error, but I cannot edit it anymore. Sorry. The table should read like ``` Formula => ELO = a * DWZ + b Class n Mean DWZ Mean Elo Slope (a) Intercept (b) Pearson r R^2 approx. DWZ Range A 602 2135.6495 2139.1362 0.932880 146.831335 0.959376 0.920403 2000-2500 B 1010 1660.6950 1762.4703 0.514670 907.759979 0.839404 0.704600 1400-2000 C 106 1166.7830 1513.0849 0.159700 1326.749467 0.446591 0.199443 1000-1600 Inverted => DWZ = a' * ELO + b' Class Slope(a') Intercept(b') A 1.072146 -157.460 B 1.943044 -1765.019 C 6.262484 -8306.006 ```